A comparison of some optimisation techinques

  • Optimisation and regression are related in that they seek a solution to a supervised problem that involves an error measure as the guide to the eventual solution.

    In this article, Paul Beinat who leads Finity's AI practice area explores optimisation techniques that are used where regression gives poor results, or where it is not possible to use it. He examines the performance of two optimisation variants based on the Nelder-Mead simplex and a particle swarm. To make empirical comparisons we use two complex error functions that have local minima.